
Abstract— This paper presents a network-based model
predictive controller design strategy for spatially distributed
systems (SDSs), where control input quantization over a digital
network is considered. To simplify the control design, the
spatiotemporal control problem is transformed to a temporal
one. Basing on the temporal model, a robust model predictive
controller is designed to provides an efficient solution to the
quantized control issues for spatially distributed system, where
the quantization problem for control design is solved by a robust
control problem with sector bounds uncertainties. The efficiency
of the proposed method was shown through a network control
problem based on a typical SDS.
I. INTRODUCTION
Over the past decade, most literature on network-based
control strategies focuses on lumped parameter systems
(LPSs) that has gained a lot of useful conclusions [1-3].
Basically, most researches focus on the control problem
bought by communication delays, different sampling intervals,
or quantization problem. As the communication network has
great advantages, it can be used not only for LPSs, but also for
SDSs. These SDSs are typically for those plants that can be
modeled by diffusion PDEs, such as chemical reactors through
long distance control, etc. [4,5]. Thus, it is necessary to
integrate and address the control problems for spatially
distributed systems (SDSs) under new circumstances brought
by communication network Recently, researches on network
based control for SDSs have gradually gained some useful
conclusions [6-13]. In [9] and [13], estimation methods and
control design are considered for SDSs based on the sensor
networks. In [7,10,11,12], different networked controller, such
as networked H-infinity, networked MPC are proposed for
SDSs under network communication circumstance.
For most network based control strategy for SDS, it has to
deal with two aspects problem. One is how to obtain the
approximated finite dimensional models from the partial
differential equations (PDEs), since it is difficult to consider a
control strategy based on PDEs directly [14]. The other is how
to deal with the control problems for SDS brought by
communication network. And most conclusions for time-delay
or drop-package problems are derived based on the
assumption that the bandwidth of the network is infinite. Thus,
*This work was supported by National Nature Science Foundation of
China No. 61673177, No. 61673173, No. 61673178, No. 61603205, the
Natural Science Foundation of Shanghai No. 16ZR1407400, the
Fundamental Research Funds for the Central Universities (No.
22A201514048).
M.L. Wang, H.C. Yan, and H.B. Shi are with the Key Laboratory of
Advanced Control and Optimization for Chemical Processes of Ministry of
Education, East China University of Science and Technology, 130, Meilong
Road, Shanghai 200237, China (e-mail: wml_ling@ ecust.edu.cn).
B. Q. Xue is with the Qingdao University, Qingdao, China.
in this paper, the control problem is designed for SDSs,
considering the data is communicated through a quantizer with
capacity-limited feedback signal.
Meanwhile, considering the merits of MPC strategy, this
paper proposes a robust model predictive control strategy for
SDSs to solve the quantized issues of networked control
systems, extending the robust MPC strategy proposed in [15]
for LPS to SDS. And it obtains the approximated finite
dimensional models by using a time/space separation method,
which is literally easier to calculate from input/output data
sets. First, a low dimensional temporal model is used to predict
the dynamics of the SDS based on the input/output data sets.
Based on it, the control objective is redefined in terms of the
outputs of low-dimensional time series. As the spatial
information can be described by spatial basis functions (a
defined matrix), the robust MPC strategy proposed for LPS
can be easily extended for the SDS, where the control strategy
is designed on the basis of the derived state space model type
by fully considering the quantization through the
communication network. Thus, by sending the optimizing
sequence of control input over a digital network, the quantized
feedback control problem is transformed into a robust control
problem with bounded uncertainties.
The paper is organized in five sections. In the next section,
the approximated model is presented for control design. In
Section 3, it proposed a the robust MPC strategy design for
quantized SDS. Section 4 shows the application to a spatially
distributed process. In this section, we will analyze the results
and discuss the advantages of the proposed method. In the last
section, the conclusion is drawn.
Notations: In the paper,
and
are used to denote
the n-dimensional Euclidean space and the set of all
real
matrices respectively; The notation
or
(
) explains that the matrix
is a real symmetric
positive definite matrix or positive semi-definite matrix;
represents a symmetric matrix, where
refers to a
symmetric element or submatrix, and
,
and
are
arbitrarily matrices.
II. APPROXIMATED MODEL FOR CONTROL DESIGN
Consider an industrial process, the input/output data sets
are sampled through by finite sensors and actuators. The
spatial-temporal outputs
are measured at the N
spatial locations
and the temporal signal
are
obtained from finite actuators.
is the spatial variable
and
is spatial domain.
Network-based Predictive Control Strategy for Spatially Distributed
System Based with quantization *
M. L. Wang, B. Q. Xue, H. C. Yan, and H.B. Shi
2017 11th Asian Control Conference (ASCC)
Gold Coast Convention Centre, Australia
December 17-20, 2017
978-1-5090-1573-3/17/$31.00 ©2017 IEEE 1075